© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 20902
    Prev
    Next
    INDEX
    Explanations

    references to physical measurements, specifically focusing on height

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    torch.float32
    Hook Point
    blocks.8.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    8
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    99
    +0.13
    0.6%
    35
    +0.11
    0.5%
    288
    +0.10
    0.5%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    99
    +0.13
    0.01
    35
    +0.11
    0.01
    89
    +0.10
    0.01
    Negative Logits
    tical
    -0.77
    ño
    -0.76
    BOOK
    -0.73
    utical
    -0.71
    zsche
    -0.71
     Reloaded
    -0.70
    uterte
    -0.67
    Ò
    -0.66
    ãģ¦
    -0.66
    ctuary
    -0.66
    POSITIVE LOGITS
    ened
    0.91
     heights
    0.83
     height
    0.82
    itudinal
    0.81
    ening
    0.81
     cliffs
    0.80
     altitude
    0.79
     stature
    0.79
    rises
    0.76
     variance
    0.76
    Activations Density 0.017%

    No Known Activations